Learning AI Driven (LAD)

LAD creates a local workspace for learning any skill with an AI agent as a learning collaborator. It uses a repeatable loop: learn, recall, apply, receive feedback, and review after a delay.

Start a learning workspace

Create and enter a new empty folder, then run:

npx lad install

The installer refuses to run in a non-empty directory. It creates only files inside the directory where you run it.

Update an existing workspace

From inside a workspace created by LAD, run:

npx lad update

This adds any new LAD agent skills that are missing. It does not overwrite your learning files or existing skills, so local customizations stay intact.

What LAD creates

learning/
├── learning-plan.md       Your chosen skill, goal, and starting point
├── sessions/              One learning record per session
└── reviews/               Notes and prompts for scheduled reviews
.agents/skills/
├── learning-coach/        Instructions for an agent supporting the learning loop
└── deep-research-learning/ Researches a topic and curates a learning path and sources
README.md                  How to use this workspace

Use the learning loop

1. Choose one small concept or skill in learning/learning-plan.md.
2. Ask an agent using deep-research-learning to find and assess learning materials for it.
3. Ask an agent using the learning-coach skill to help you learn it.
4. Answer recall questions without looking at your notes.
5. Complete an application task and use the feedback to correct it.
6. Save a learning record in learning/sessions/ and schedule a review.

Completion is not proof of learning. The aim is durable transfer: independently using the skill in a relevant new situation after time has passed.

See the learning-flow guide for the methodology behind LAD.